Note that this appears to be enforced at the EventTarget level, not at the AbortController/AbortSignal level or at the Event level: https://jsfiddle.net/wL5evuh3/3/.

Given that this is about the EventTarget itself, and thus isn't covered by the Web IDL issue you list, probably the check would have to be at https://dom.spec.whatwg.org/#concept-event-listener-inner-invoke . You might think we could factor it out to surround all of https://dom.spec.whatwg.org/#concept-event-dispatch but that would probably not do the right thing if the first listener removed the frame. (In that case I assume subsequent listeners should not run... but someone should test.)

Note that this appears to be enforced at the EventTarget level, not at the AbortController/AbortSignal level or at the Event level: https://jsfiddle.net/wL5evuh3/3/.

Nice, right, this isn't specific to AbortSignal --- just using that as an example because that's just how I discovered this.

Given that this is about the EventTarget itself, and thus isn't covered by the Web IDL issue you list, probably the check would have to be at https://dom.spec.whatwg.org/#concept-event-listener-inner-invoke . You might think we could factor it out to surround all of https://dom.spec.whatwg.org/#concept-event-dispatch but that would probably not do the right thing if the first listener removed the frame. (In that case I assume subsequent listeners should not run... but someone should test.)

Re: WebIDL, I was curious what should happen if constructing an object in a realm whose global's document is detached. This happens earlier (step 2 of fire an event), but Blink doesn't create the wrapper until much later.

I'll test the "detached during an event with subsequent handlers" case. I was thinking about that as well and was planning to write a WPT for that assuming we make a spec change. I agree that the check can't happen too early because if this case.

It looks like all the engines don't run subsequent listeners after the event target's context gets detached. For Blink and WebKit, this follows from where we're checking this; for Gecko, I think the listeners get cleared on detach, so iteration over the listeners stops after detach.

And yeah, there would need to be a check in https://dom.spec.whatwg.org/#concept-event-listener-inner-invoke to account for detaching from a listener. I'll try to send a PR next week along with some tests.

cc @smaug----